Horizontal Axis Wind Turbines (HAWTs) are a very popular source of renewable energy, but these machines are often subject to off-design conditions owing to the unsteady nature of wind. Unsteady aerodynamic conditions have been associated with cyclic loading of turbine blades, which can have hugely detrimental effects on their operational life. When subjected to unsteady aerodynamic conditions, the blade sees an overshoot of lift as compared to steady cases. At high angles of attack, the blade witnesses a very sharp stall from which it may or may not recover. The excursion of lift and subsequent sharp stall cause the lift curve to exhibit hysteresis behaviour, which outlines the phenomenon of dynamic stall. Continued recurrence of dynamic stall in gusty conditions leads to increased risk of failure due to fatigue loading. The primary aim of this project was to investigate the prediction of dynamic stall in quasi two-dimensional (2D) and strongly three-dimensional (3D) flows, using the Lattice-Boltzmann Method (LBM) based solver, PowerFLOW. The first part of the project dealt with simulation of dynamic stall in quasi-two dimensional flow. Validation of the computational data was performed against experimental data published by the Ohio State University (OSU) and the National Renewable Energy laboratory (NREL). Simulations were performed for two configurations -- the first set with a clean leading edge, and the second set employing a leading edge surface roughness. The second part of the project dealt with the analysis of strongly 3D flow conditions wherein the pitching blade was also subject to rotation. Rotation was implemented using a stationary local reference frame. In both stages, flow field data was analysed using a mix of velocity, vorticity and pressure plots. For the 3D simulation, skin friction was also investigated as a measure of flow separation. In context of the quasi-2D cases, it was found that PowerFLOW faces some limitations when attempting to match experimental data. This was attributed to an incorrect prediction of flow separation, consequent of using a wall model for resolution of near-wall structures. This issue was more pronounced in the clean cases, where laminar-turbulent transition is critical. The tripping of the boundary layer at the leading edge was also seen to adversely affect the formation of the Dynamic Stall Vortex (DSV). For the 3D cases, it was seen that rotation causes an even greater lift excursion. The cause for this was attributed to the generation of strong spanwise pressure gradients, leading to significant spanwise flow. This spanwise flow was found to be present only in regions where flow separation would be expected in a quasi-2D setup. The findings of this research project helped elaborate on our current understanding of dynamic stall, in addition to highlighting some important technical considerations for LBM-based flow solution using PowerFLOW. This research stands to enable inclusion of more complex physics, and further the capabilities of modern computational fluid dynamics solvers.